Part 3 Data cleaning - Mammals
All tracking devices come with errors, and each track must be checked visual and cleaned of outliers before analysis.
This page shows this process for each mammal within the project. It has been automated so that we always have access to a stardardised and cleaned locations database.
3.0.1 Tapirs
For the tapir collars we use the Telonics - TGW-4577-4 model, wioth data sent through the Iridium system.
They have low voltage and mortality alerts.
There are two types of location data these collars collect - regular GPS and QFP. QFP’s are Quick Fix Positions and use less battery than a standard GPS fix. Units can be configured to collect GPS, QFP, or a mix of the two location types - we collect a mix.
Our collars collect QFP’s when there is no GPS signal available.
Succeded = a successful GPS location
Resolved QFP = a successful QFP location
Resolved QFP = an uncertain QFP location (lower quality)
Unresolved QFP = A poor quality QFP
Lets explore the implications of these different fix types for one individual - Juancito.
3.0.1.1 Juancito
## Reading layer `OC_trails2' from data source
## `/Users/runner/work/Osa-Conservation-Movement-Ecology-Lifetime-Exploration/Osa-Conservation-Movement-Ecology-Lifetime-Exploration/shapefiles/OC_trails2.shp'
## using driver `ESRI Shapefile'
## Simple feature collection with 204 features and 1 field
## Geometry type: MULTILINESTRING
## Dimension: XY
## Bounding box: xmin: -83.35712 ymin: 8.393359 xmax: -83.30851 ymax: 8.429462
## Geodetic CRS: WGS 84First observation: 2022-05-29 18:01:25
Last observation: 2023-11-09 04:02:20
Active for: 528.4 days
To date we have 4011 fixes from Juancito. Of those, they are of the following types.
| Var1 | Freq | Percentage |
|---|---|---|
| Resolved QFP | 2380 | 0.59 |
| Resolved QFP (Uncertain) | 80 | 0.02 |
| Succeeded | 537 | 0.13 |
| Unresolved QFP | 1014 | 0.25 |
Plot of the raw data:
We can then explore how these different fix types introduce error into the locations:
NOTE ZOOM IN!
So there is unreliability in all of the QFP estimates, but the GPS look great (not much spread).
Lets explore the error around each type of fix (if it has an estimate present):
Only GPS data has an error in meters (max 127m):
QFP data has error in doplar estimates:
So we have a huge spread in error. Lets subset to anything with a hdop <10,and resolved. This substantially reduced the spread but still leaves a couple of issues - notable outliers.
We should also use ctmm (as in the vultures) to find errors:
3.0.1.2 Tuva
First observation: 2023-04-14 08:01:55
Last observation: 2024-05-17 04:00:37
Active for: 398.8 days
To date we have 2670 fixes from Tuva. Of those, they are of the following types.
| Var1 | Freq | Percentage |
|---|---|---|
| Resolved QFP | 1402 | 0.53 |
| Resolved QFP (Uncertain) | 56 | 0.02 |
| Succeeded | 307 | 0.11 |
| Unresolved QFP | 905 | 0.34 |
Which means even less “resolved” locations.
Plot of the raw data:
We can then explore how these different fix types introduce error into the locations:
NOTE ZOOM IN!
So there is unreliability in all of the QFP estimates, but the GPS look great (not much spread).
Lets explore the error around each type of fix (if it has an estimate present):
Only GPS data has an error in meters (max 127m):
QFP data has error in dopler estimates:
So we have a huge spread in error. Lets subset to anything with a hdop <10,and resolved. This substantially reduced the spread but still leaves a couple of issues.
We should also use ctmm (as in the vultures) to find errors:
3.0.2 Carnivores
For carnivores we use Vectronic collars.
Large carnivores we have VERTEX Plus (GPS + ACC) and VERTEX Vision (GPS + ACC + Video)
3.0.2.1 Killa
VERTEX Plus collar.
First observation: 2023-10-31 14:00:46
Last observation: 2024-03-09 19:00:46
Active for: 130.2 days
To date we have 672 fixes from . Of those, they are of the following types.
| Var1 | Freq | Percentage |
|---|---|---|
| 3D GPS Fix | 19 | 0.03 |
| 3D valid GPS Fix | 653 | 0.97 |
Plot of the raw data:
We can then explore how these different fix types introduce error into the locations:
NOTE ZOOM IN!
We get doppler error estimates on these data, and they are all fairly small!:
The average number of fixes per day to date 5.2 per day.
Those are distributed through time as follows:
No cleaning done to date:
We should also use ctmm (as in the vultures) to find errors:
3.0.2.2 Tico
VERTEX Plus collar.
First observation: 2024-04-04 07:01:53
Last observation: 2024-05-26 02:00:56
Active for: 51.8 days
To date we have 667 fixes from . Of those, they are of the following types.
| Var1 | Freq | Percentage |
|---|---|---|
| 3D GPS Fix | 2 | 0 |
| 3D valid GPS Fix | 665 | 1 |
Plot of the raw data:
We get doppler error estimates on these data, and they are all fairly small!:
The average number of fixes per day to date 12.9 per day.
Those are distributed through time as follows:
No cleaning done to date:
We should also use ctmm (as in the vultures) to find errors:
Current battery level:
Be worried if it goes below 3000!
3.0.2.3 Coyunda
VERTEX Plus collar.
First observation: 2024-04-15 14:01:16
Last observation: 2024-05-21 05:00:14
Active for: 35.6 days
To date we have 464 fixes from . Of those, they are of the following types.
| Var1 | Freq | Percentage |
|---|---|---|
| 3D GPS Fix | 3 | 0.01 |
| 3D valid GPS Fix | 461 | 0.99 |
Plot of the raw data:
We get doppler error estimates on these data, and they are all fairly small!:
The average number of fixes per day to date 13 per day.
Those are distributed through time as follows:
No cleaning done to date:
We should also use ctmm (as in the vultures) to find errors:
Current battery level:
Be worried if it goes below 3000!